1 /*
2 * Bink Audio decoder
3 * Copyright (c) 2007-2011 Peter Ross (pross@xvid.org)
4 * Copyright (c) 2009 Daniel Verkamp (daniel@drv.nu)
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 /**
24 * @file
25 * Bink Audio decoder
26 *
27 * Technical details here:
28 * http://wiki.multimedia.cx/index.php?title=Bink_Audio
29 */
30
31 #include "config_components.h"
32
33 #include "libavutil/channel_layout.h"
34 #include "libavutil/intfloat.h"
35 #include "libavutil/mem_internal.h"
36
37 #define BITSTREAM_READER_LE
38 #include "avcodec.h"
39 #include "dct.h"
40 #include "decode.h"
41 #include "get_bits.h"
42 #include "codec_internal.h"
43 #include "internal.h"
44 #include "rdft.h"
45 #include "wma_freqs.h"
46
47 #define MAX_DCT_CHANNELS 6
48 #define MAX_CHANNELS 2
49 #define BINK_BLOCK_MAX_SIZE (MAX_CHANNELS << 11)
50
51 typedef struct BinkAudioContext {
52 GetBitContext gb;
53 int version_b; ///< Bink version 'b'
54 int first;
55 int channels;
56 int ch_offset;
57 int frame_len; ///< transform size (samples)
58 int overlap_len; ///< overlap size (samples)
59 int block_size;
60 int num_bands;
61 float root;
62 unsigned int bands[26];
63 float previous[MAX_DCT_CHANNELS][BINK_BLOCK_MAX_SIZE / 16]; ///< coeffs from previous audio block
64 float quant_table[96];
65 AVPacket *pkt;
66 union {
67 RDFTContext rdft;
68 DCTContext dct;
69 } trans;
70 } BinkAudioContext;
71
72
decode_init(AVCodecContext * avctx)73 static av_cold int decode_init(AVCodecContext *avctx)
74 {
75 BinkAudioContext *s = avctx->priv_data;
76 int sample_rate = avctx->sample_rate;
77 int sample_rate_half;
78 int i, ret;
79 int frame_len_bits;
80 int max_channels = avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT ? MAX_CHANNELS : MAX_DCT_CHANNELS;
81 int channels = avctx->ch_layout.nb_channels;
82
83 /* determine frame length */
84 if (avctx->sample_rate < 22050) {
85 frame_len_bits = 9;
86 } else if (avctx->sample_rate < 44100) {
87 frame_len_bits = 10;
88 } else {
89 frame_len_bits = 11;
90 }
91
92 if (channels < 1 || channels > max_channels) {
93 av_log(avctx, AV_LOG_ERROR, "invalid number of channels: %d\n", channels);
94 return AVERROR_INVALIDDATA;
95 }
96 av_channel_layout_uninit(&avctx->ch_layout);
97 av_channel_layout_default(&avctx->ch_layout, channels);
98
99 s->version_b = avctx->extradata_size >= 4 && avctx->extradata[3] == 'b';
100
101 if (avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT) {
102 // audio is already interleaved for the RDFT format variant
103 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
104 if (sample_rate > INT_MAX / channels)
105 return AVERROR_INVALIDDATA;
106 sample_rate *= channels;
107 s->channels = 1;
108 if (!s->version_b)
109 frame_len_bits += av_log2(channels);
110 } else {
111 s->channels = channels;
112 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
113 }
114
115 s->frame_len = 1 << frame_len_bits;
116 s->overlap_len = s->frame_len / 16;
117 s->block_size = (s->frame_len - s->overlap_len) * FFMIN(MAX_CHANNELS, s->channels);
118 sample_rate_half = (sample_rate + 1LL) / 2;
119 if (avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
120 s->root = 2.0 / (sqrt(s->frame_len) * 32768.0);
121 else
122 s->root = s->frame_len / (sqrt(s->frame_len) * 32768.0);
123 for (i = 0; i < 96; i++) {
124 /* constant is result of 0.066399999/log10(M_E) */
125 s->quant_table[i] = expf(i * 0.15289164787221953823f) * s->root;
126 }
127
128 /* calculate number of bands */
129 for (s->num_bands = 1; s->num_bands < 25; s->num_bands++)
130 if (sample_rate_half <= ff_wma_critical_freqs[s->num_bands - 1])
131 break;
132
133 /* populate bands data */
134 s->bands[0] = 2;
135 for (i = 1; i < s->num_bands; i++)
136 s->bands[i] = (ff_wma_critical_freqs[i - 1] * s->frame_len / sample_rate_half) & ~1;
137 s->bands[s->num_bands] = s->frame_len;
138
139 s->first = 1;
140
141 if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
142 ret = ff_rdft_init(&s->trans.rdft, frame_len_bits, DFT_C2R);
143 else if (CONFIG_BINKAUDIO_DCT_DECODER)
144 ret = ff_dct_init(&s->trans.dct, frame_len_bits, DCT_III);
145 else
146 av_assert0(0);
147 if (ret < 0)
148 return ret;
149
150 s->pkt = avctx->internal->in_pkt;
151
152 return 0;
153 }
154
get_float(GetBitContext * gb)155 static float get_float(GetBitContext *gb)
156 {
157 int power = get_bits(gb, 5);
158 float f = ldexpf(get_bits(gb, 23), power - 23);
159 if (get_bits1(gb))
160 f = -f;
161 return f;
162 }
163
164 static const uint8_t rle_length_tab[16] = {
165 2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 14, 15, 16, 32, 64
166 };
167
168 /**
169 * Decode Bink Audio block
170 * @param[out] out Output buffer (must contain s->block_size elements)
171 * @return 0 on success, negative error code on failure
172 */
decode_block(BinkAudioContext * s,float ** out,int use_dct,int channels,int ch_offset)173 static int decode_block(BinkAudioContext *s, float **out, int use_dct,
174 int channels, int ch_offset)
175 {
176 int ch, i, j, k;
177 float q, quant[25];
178 int width, coeff;
179 GetBitContext *gb = &s->gb;
180
181 if (use_dct)
182 skip_bits(gb, 2);
183
184 for (ch = 0; ch < channels; ch++) {
185 FFTSample *coeffs = out[ch + ch_offset];
186
187 if (s->version_b) {
188 if (get_bits_left(gb) < 64)
189 return AVERROR_INVALIDDATA;
190 coeffs[0] = av_int2float(get_bits_long(gb, 32)) * s->root;
191 coeffs[1] = av_int2float(get_bits_long(gb, 32)) * s->root;
192 } else {
193 if (get_bits_left(gb) < 58)
194 return AVERROR_INVALIDDATA;
195 coeffs[0] = get_float(gb) * s->root;
196 coeffs[1] = get_float(gb) * s->root;
197 }
198
199 if (get_bits_left(gb) < s->num_bands * 8)
200 return AVERROR_INVALIDDATA;
201 for (i = 0; i < s->num_bands; i++) {
202 int value = get_bits(gb, 8);
203 quant[i] = s->quant_table[FFMIN(value, 95)];
204 }
205
206 k = 0;
207 q = quant[0];
208
209 // parse coefficients
210 i = 2;
211 while (i < s->frame_len) {
212 if (s->version_b) {
213 j = i + 16;
214 } else {
215 int v = get_bits1(gb);
216 if (v) {
217 v = get_bits(gb, 4);
218 j = i + rle_length_tab[v] * 8;
219 } else {
220 j = i + 8;
221 }
222 }
223
224 j = FFMIN(j, s->frame_len);
225
226 width = get_bits(gb, 4);
227 if (width == 0) {
228 memset(coeffs + i, 0, (j - i) * sizeof(*coeffs));
229 i = j;
230 while (s->bands[k] < i)
231 q = quant[k++];
232 } else {
233 while (i < j) {
234 if (s->bands[k] == i)
235 q = quant[k++];
236 coeff = get_bits(gb, width);
237 if (coeff) {
238 int v;
239 v = get_bits1(gb);
240 if (v)
241 coeffs[i] = -q * coeff;
242 else
243 coeffs[i] = q * coeff;
244 } else {
245 coeffs[i] = 0.0f;
246 }
247 i++;
248 }
249 }
250 }
251
252 if (CONFIG_BINKAUDIO_DCT_DECODER && use_dct) {
253 coeffs[0] /= 0.5;
254 s->trans.dct.dct_calc(&s->trans.dct, coeffs);
255 }
256 else if (CONFIG_BINKAUDIO_RDFT_DECODER)
257 s->trans.rdft.rdft_calc(&s->trans.rdft, coeffs);
258 }
259
260 for (ch = 0; ch < channels; ch++) {
261 int j;
262 int count = s->overlap_len * channels;
263 if (!s->first) {
264 j = ch;
265 for (i = 0; i < s->overlap_len; i++, j += channels)
266 out[ch + ch_offset][i] = (s->previous[ch + ch_offset][i] * (count - j) +
267 out[ch + ch_offset][i] * j) / count;
268 }
269 memcpy(s->previous[ch + ch_offset], &out[ch + ch_offset][s->frame_len - s->overlap_len],
270 s->overlap_len * sizeof(*s->previous[ch + ch_offset]));
271 }
272
273 s->first = 0;
274
275 return 0;
276 }
277
decode_end(AVCodecContext * avctx)278 static av_cold int decode_end(AVCodecContext *avctx)
279 {
280 BinkAudioContext * s = avctx->priv_data;
281 if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
282 ff_rdft_end(&s->trans.rdft);
283 else if (CONFIG_BINKAUDIO_DCT_DECODER)
284 ff_dct_end(&s->trans.dct);
285
286 return 0;
287 }
288
get_bits_align32(GetBitContext * s)289 static void get_bits_align32(GetBitContext *s)
290 {
291 int n = (-get_bits_count(s)) & 31;
292 if (n) skip_bits(s, n);
293 }
294
binkaudio_receive_frame(AVCodecContext * avctx,AVFrame * frame)295 static int binkaudio_receive_frame(AVCodecContext *avctx, AVFrame *frame)
296 {
297 BinkAudioContext *s = avctx->priv_data;
298 GetBitContext *gb = &s->gb;
299 int ret;
300
301 again:
302 if (!s->pkt->data) {
303 ret = ff_decode_get_packet(avctx, s->pkt);
304 if (ret < 0) {
305 s->ch_offset = 0;
306 return ret;
307 }
308
309 if (s->pkt->size < 4) {
310 av_log(avctx, AV_LOG_ERROR, "Packet is too small\n");
311 ret = AVERROR_INVALIDDATA;
312 goto fail;
313 }
314
315 ret = init_get_bits8(gb, s->pkt->data, s->pkt->size);
316 if (ret < 0)
317 goto fail;
318
319 /* skip reported size */
320 skip_bits_long(gb, 32);
321 }
322
323 /* get output buffer */
324 if (s->ch_offset == 0) {
325 frame->nb_samples = s->frame_len;
326 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
327 return ret;
328 }
329
330 if (decode_block(s, (float **)frame->extended_data,
331 avctx->codec->id == AV_CODEC_ID_BINKAUDIO_DCT,
332 FFMIN(MAX_CHANNELS, s->channels - s->ch_offset), s->ch_offset)) {
333 av_log(avctx, AV_LOG_ERROR, "Incomplete packet\n");
334 s->ch_offset = 0;
335 return AVERROR_INVALIDDATA;
336 }
337 s->ch_offset += MAX_CHANNELS;
338 get_bits_align32(gb);
339 if (!get_bits_left(gb)) {
340 memset(gb, 0, sizeof(*gb));
341 av_packet_unref(s->pkt);
342 }
343 if (s->ch_offset >= s->channels) {
344 s->ch_offset = 0;
345 } else {
346 goto again;
347 }
348
349 frame->nb_samples = s->block_size / FFMIN(avctx->ch_layout.nb_channels, MAX_CHANNELS);
350
351 return 0;
352 fail:
353 s->ch_offset = 0;
354 av_packet_unref(s->pkt);
355 return ret;
356 }
357
decode_flush(AVCodecContext * avctx)358 static void decode_flush(AVCodecContext *avctx)
359 {
360 BinkAudioContext *const s = avctx->priv_data;
361
362 /* s->pkt coincides with avctx->internal->in_pkt
363 * and is unreferenced generically when flushing. */
364 s->first = 1;
365 s->ch_offset = 0;
366 }
367
368 const FFCodec ff_binkaudio_rdft_decoder = {
369 .p.name = "binkaudio_rdft",
370 .p.long_name = NULL_IF_CONFIG_SMALL("Bink Audio (RDFT)"),
371 .p.type = AVMEDIA_TYPE_AUDIO,
372 .p.id = AV_CODEC_ID_BINKAUDIO_RDFT,
373 .priv_data_size = sizeof(BinkAudioContext),
374 .init = decode_init,
375 .flush = decode_flush,
376 .close = decode_end,
377 FF_CODEC_RECEIVE_FRAME_CB(binkaudio_receive_frame),
378 .p.capabilities = AV_CODEC_CAP_DR1,
379 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE | FF_CODEC_CAP_INIT_CLEANUP,
380 };
381
382 const FFCodec ff_binkaudio_dct_decoder = {
383 .p.name = "binkaudio_dct",
384 .p.long_name = NULL_IF_CONFIG_SMALL("Bink Audio (DCT)"),
385 .p.type = AVMEDIA_TYPE_AUDIO,
386 .p.id = AV_CODEC_ID_BINKAUDIO_DCT,
387 .priv_data_size = sizeof(BinkAudioContext),
388 .init = decode_init,
389 .flush = decode_flush,
390 .close = decode_end,
391 FF_CODEC_RECEIVE_FRAME_CB(binkaudio_receive_frame),
392 .p.capabilities = AV_CODEC_CAP_DR1,
393 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE | FF_CODEC_CAP_INIT_CLEANUP,
394 };
395